As an imaging system becomes more effective at mimicking human eyesight, it generally grows more powerful and versatile. This can be seen throughout machine vision, especially in the emergence of 3D stereoscopic imaging systems based on the eye.
Computational imaging – replacing optical techniques with computational ones – is the key.
Although computational imaging is a young technology, it offers tremendous potential. Computational techniques are so robust and efficient they can record high-resolution visual information – even entire movies – using one pixel.
Other breakthroughs include:
- Lenseless cameras;
- 3D imaging systems;
- “Foveated” imaging.
What Is “Foveation?” A Single Pixel Can Mimic Animal Sight
Many animals, including humans, have foveated vision. This means in any given instant, details near the visual field's center are seen with the greatest acuity. Details in the periphery are perceived less sharply.
Recently, a team led by David Phillips at the University of Glasgow developed sophisticated imaging techniques relying on a single pixel implementation of this concept. They even figured out how to move the foveated region to follow objects in the user’s field of view.
An Imaging System That Records Motion, Inspired by the Eye
Although Phillips and his colleagues have refined the implementation of single pixel imaging, they weren’t the first ones to take fresh inspiration from the human vision system. As early as 2013, the Swiss company iniLabs made strides in this area.
The iniLabs Dynamic Vision Sensor (DVS) 128 camera isn’t a standard sensor: It takes a cue from the human retina by recording only motion in its field of view rather than all the visual information. This lets it efficiently capture specialized data for long running times.
Imaging systems continue to adapt key principles from their organic counterparts. The next step in computational imaging could blend biology and technology in totally new ways.